Predictive Maintenance programs to increase equipment productivity are a fashion trend of the current time. However, doubts about this approach are growing. Having the equipment and process data centrally available is undoubtedly correct. However, there is a better and broader approach to take advantage of investing in information technology.
The promise of Predictive Maintenance is tempting: With machine learning technologies searching in historical performance, standstill and sensor data to find connections between faults and causes, in order to predict when a system component will fail, in order to proactively exchange it shortly before this defect. The aim is to transform unplanned downtimes into planned downtimes and thus increase equipment productivity.
At first glance, the procedure appears to be a worthwhile innovation in almost all manufacturing machines. But can this promise be kept?
If you analyze the situation more deeply, the optimization potential through predictive maintenance is reduced depending on 4 factors:
- Too few sensors: It is only possible to predict what is also available from detailed data. However, there may be many technical defects on components that are not subject to sensor monitoring.
- Too little data: Predicting defects is more difficult than you might think. Defects in individual system parts are very rare events, which means that there is typically not enough data to train a reliable model of machine learning.
- Too little effect: System defects reduce the OEE, but physical or planning redundancy, a clever spare parts strategy and quickly available and qualified personnel can keep the downtime of a system short in the event of a technical defect.
- Too strong focus: If you look at a distribution of all equipment losses, you can see not only the technical defects but also further availability losses such as organizational disruptions (no personnel, no material, waiting for Q approval …), performance losses such as slow production speed or quality losses as times of suboptimal productivity. As a rule, these losses represent the clear majority of productivity losses, so that the importance of technical defects takes a back seat.
For one or more of these four reasons, expectations of predictive maintenance projects often fall short of reality.
Does this mean that the manufacturing industry is not a field of application for data-driven productivity initiatives? No, just the opposite. However, the narrow focus on predicting future technical defects is usually* not the right approach. Instead, the comprehensive approach of analyzing and optimizing OEE losses in terms of targeted OEE management seems promising.
For this, the causes of the loss in terms of availability, performance and quality the deviation from the ideal run of the equipment needs to be classified. A small part of the data can be supplied from the system control, but the majority of the classifications result from the domain knowledge of the operator. The following figure shows the structure:
With a transparent reporting system on the data collected in this way, a large contribution can be made to focusing improvement activities and thus increasing equipment productivity. In addition, artificial intelligence, which is usually machine learning, or statistics in this data stream can be used to gain interesting insights into patterns and anomalies for the possibilities of increasing equipment productivity:
- Grouping frequency / duration of causes of loss
- Identification of periods with accumulation / absence of losses
- Identification of abnormal (e.g. especially long or short) losses
- Identification of cyclical availability and performance losses
- Determination of the set-up quality of a conversion
- Calculation of a key figure for the running stability or identification of periods of stable or unstable system operation
- Identification of a temporary deviation from the longer-term defect trends
If selected analyzes of these are carried out on-line, it is even possible to intervene in the operation while the fault still exists. For example, an unusually long standstill according to the evaluation of the algorithm can be identified and reported to a manager so that he or she can get a personal impression and initiate measures depending on the situation.
If necessary, this can be enriched with sensor data that is usually recorded for Predictive Maintenance: vibrations, temperature, pressures or the like in connection with loss of availability, performance and quality. However, this data is only a supplement to the further refinement of the methodology. So the cherry on the cake rather than the cake itself.
All analyzes have in common that the technology gives an indication of the possibility of optimization, but humans have to make decisions based on this data and act in a targeted manner.
The potential of data-driven process improvement goes far beyond Predictive Maintenance. For manufacturing companies, OEE management offers digital optimization approaches that are easier to implement and offer greater added value. So why limit your options with the pure focus on Predictive Maintenance?
* To avoid misunderstandings: Of course there are also worthwhile applications for predictive maintenance technologies e.g. in safety-relevant environments (aircraft engines) or in the event of downtimes with enormous follow-up costs (off-shore wind turbines, chemical plants).